167 SOATO-ENRITIC INTERACTIONS UNERLYING ACTION POTENTIAL GENERATION IN NEOCORTICAL PYRAIAL CELLS IN VIVO Alain estexhe, 1 Eric J. Lang 2 and enis Pare 1 1 Laboratoire de Neurophysiologie, Universite Laval, Quebec G1K 7P4, Canada 2 epartment ofphysiology and Neuroscience, New York University, New York, NY 10016, USA In: Computational Neuroscience: Trends in Research, 1998, Edited by Bower, J., Plenum Press, New York, pp. 167-172, 1998. Action potential (AP) generation and propagation through the dendritic tree of pyramidal neurons has received much attention lately 1. Experiments indicate that inhibitory postsynaptic potentials (IPSPs) have a decisive eect in controlling action potential invasion in dendrites 2,3.However, these phenomena were observed in slices, and remain to be investigated under in vivo conditions where neurons are subject to intense synaptic activity. Here, we have combined computational models with intracellular recordings of neocortical pyramidal cells in order to investigate how spikes can be controlled by IPSPs in vivo. EXPERIENTAL OBSERVATIONS We obtained intracellular recordings of morphologically-identied pyramidal neurons (n=42) from the parietal cortex (areas 5-7) in barbiturate-anesthetized cats and compared APs evoked by synaptic inputs, antidromic invasion, or direct current injection (methods were described in detail elsewhere 4 ). The neurons were recorded within 1-2 mm of a ten-electrode array (Fig. 1A). Evidence was obtained that cortical stimuli delivered at dierent depths activate partially segregated sets of aerents that preferentially end at corresponding cortical depths, thus exerting maximal eects on dierent compartments of pyramidal neurons. As shown in Fig. 1B, IPSPs elicited by deep cortical shocks were larger in amplitude (12.4 2.25 mv compared to 4.8 1.89 mv), had a shorter peak-latency (23.4 2.93 ms compared to 41.3 5.33 ms), a more positive reversal potential (-73.8 1.64 mv compared to -80.3 1.75 mv) and were associated with larger decreases in
Figure 1. Spike attenuation in neocortical pyramidal cells in vivo. A. Experimental paradigm. B. EPSP/IPSP sequences for proximal (1-2) and supercial (9-10) microstimulation. C. Spike attenuation with K-Acetate-lled electrodes (irect = spike evoked by current injection).. Spikes evoked by current injection. 1: control, 2: in the presence of evoked IPSP, 3: progressive attenuation of amplitude with latency; 4: comparison between evoked and spontaneous IPSPs. E. Spike attenuation with Chloride-lled electrodes. odied from a previous study 4. 168
169 input resistance (63 8.4% compared to 27 5.1%) than those elicited by supercial cortical stimuli. These dierences were statistically signicant (paired t-test, p<0.05). Using this paradigm, we investigated the eect of synaptic activity occurring in dierent somatodendritic compartments on the shape of the somatic action potential. The intensity of the cortical stimuli was adjusted to just above the spike threshold from rest. In pyramidal cells recorded with KAc pipettes (n=29), orthodromic spikes were reduced in amplitude and duration (Fig. 1C2-3) compared to APs elicited by short depolarizing current pulses (Fig. 1C1), consistent with previous ndings in the amygdala 5. The magnitude of the amplitude decrement was a function of the cortical stimulation depth relative to the position of the recorded cells (Fig. 1C). In both deep and supercial cells, maximal spike reductions were obtained with stimuli applied at the soma level. To ensure that these dierences between current-evoked and orthodromic spikes did not reect dierences in the slopes of the depolarization triggering the APs, we investigated the inuence of sub-threshold cortical shocks on APs evoked by intracellular current pulses adjusted to elicit spikes in 50% of the trials. As shown in Fig. 1, the amplitude of the direct spikes remained constant (Fig. 11), whereas the amplitude of spikes aected by IPSPs was progressively decreased (by up to 8 mv) as their latency increased (Fig. 12). To assess the relative importance of local changes in input resistance and membrane potential in these phenomena, pyramidal neurons were recorded with KCl pipettes (n=30). Whereas in cells recorded with KAc pipettes the IPSP reversals averaged - 78.1 0.77 mv (n=4), in neurons recorded with KCl pipettes they averaged -52 2.89 mv (n=13). Chloride diusion inside the cells thus caused a 20-25 mv shift in the IPSP reversal potential, bringing it close to the spike threshold. In these conditions, high intensity cortical shocks applied at the soma level produced spike amplitude reductions of 7.45 1.22 mv (n=10; Fig. 1E) compared to 15.58 1.24 (n=9) in cells recorded with KAc pipettes (Fig. 1C). COPUTATIONAL OELS Computational models were based on a cellular reconstruction provided by R. ouglas and K. artin (Fig 2A) and were simulated using NEURON (methods were described in detail elsewhere 4 ). The cell was corrected for spines and an axon hillock and initial segment were included. Passive parameters were estimated by tting the model to a voltage trace from a layer V cortical pyramidal cell recorded in the present series of experiments (Fig 2B). Active currents were inserted into the soma, dendrites and axon with dierent densities in accordance with available experimental evidence 6. Na + channel density was low in soma and dendrites (70 ps/m 2 as in adult pyramidal cells 6 { range tested 20-100 ps/m 2 ), but was high in the axon, as indicated by experimental 7 and modeling studies 8,9. The delayed-rectier was also distributed uniformly in the dendrites and no calcium currents were included. All currents were described by Hodgkin-Huxley equations, and two dierent kinetic models were tested for Na + and K + channels 8,10 with negligible inuence on the present results. The relative density of glutamatergic and GABAergic synapses in dierent regions of the cell was distributed based on morphological data reviewed previously 11, 12. Glutamatergic synapses were located exclusively in dendrites at more than 40 m away from the soma. GABAergic synapses were located in all compartments of the neuron, with the highest density in the somatic region 11, 12. Synaptic currents were simulated using kinetic models of postsynaptic receptors 13. The depth-dependent features of
170 A orphology B Passive response C Synaptic stimulation 70 mv Cell odel 10 mv 20 ms 70 mv Proximal 10 mv 70 mv istal 10 ms 100 µm S P v Control S P 2 ms latency 0.4 ms latency S P S P 20 mv 2 ms 100 mv S P 500 µm E Control IPSP rev -55 mv F Control No dendritic Na IPSP rev -80 mv No dendritic Na IPSP rev -55 mv 20 mv 1 ms Figure 2. odel of spike attenuation in cortical pyramidal cells. A. orphology of a reconstructed layer V pyramidal cell. B. Passive response of the model. C. EPSP/IPSP sequences evoked by proximal and distal stimulation.. Attenuation of spikes by IPSPs. The spike evoked by current injection is shown in control conditions and with IPSP at 2 dierent latencies. Bottom graphs in show the spatial prole of membrane potential (path indicated by dotted line in A). These plots were made over a period of 12 ms in steps of 0.2 ms (from top to bottom). E. Eect of IPSP reversal on spike amplitude. F. Eect of removing dendritic Na + channels on spike amplitude. odied from a previous study 4.
171 A Control (no IPSP) B 1.28 mv IPSP C 1.87 mv IPSP 2.42 mv IPSP E F G 4.27 mv IPSP 6.84 mv IPSP 9.61 mv IPSP 10.41 mv IPSP H 100 mv 500 µm Figure 3. Simulated IPSPs are very eective in suppressing backpropagating action potentials. Each plot represents dendritic membrane potential proles similar to Fig. 2 (0.4 ms latency). A-H: IPSPs of increasing somatic amplitude (indicated in each graph) inuenced the velocity of backpropagating spikes (B-C), then suppressed backpropagation (-H). odied from a previous study 4. evoked EPSP/IPSP sequences (Fig 2C) could be reproduced assuming that stimuli activated synapses preferentially in a relatively localized dendritic region 4. In the absence of other stimuli, spikes evoked by current injection initiated in the initial segment and backpropagated successively to soma and dendrites (Fig 2, Control). The amplitude and duration of action potentials was markedly aected by IPSPs. IPSPs occurring with small latencies with respect to the spike led to signicant reductions up to 25 mv (Fig 2), with maximal attenuations for IPSPs simulated at the perisomatic level. To simulate the eect of Chloride injection (Fig. 1E), the IPSP reversal potential was changed to -55 mv. The signicant attenuation observed in these conditions (Fig 2E) shows that the increased conductance due to IPSPs (shunt) only accounts for about 60% of spike attenuation. Removing dendritic Na + channels (Fig 2F) accounted for the remaining 40 % of spike attenuation, showing that dendritic Na + currents signicantly contribute to action potential amplitude. The full attenuation of spike amplitude could be accounted by combining this eect with the shunt described above (Fig 2F, lowermost curve). The model therefore suggests that spike amplitudes are reduced proportionally to the dendritic area that is \shut-o" by IPSPs, resulting in a spike that contains less or no participation from dendritic Na + channels. In addition, there is an important current leak through the open channels underlying the IPSPs, resulting in further attenuation. These two factors diminish in importance with more supercial stimuli. A corollary observation is that, by preventing the participation of dendritic Na + channels, IPSPs suppress spike backpropagation in distal dendrites (Fig 3). Backpropagating action potentials are fragile 2, 3, 14, 15 and the present simulations show that they can be suppressed by small-amplitude IPSPs. As these IPSPs are well within the range of spontaneous IPSPs occurring in vivo (see Fig 14), our results corroborate recent optical imaging evidence that spike-related calcium transients remains conned
172 to proximal dendrites in vivo 16. CONCLUING REARKS In conclusion, by combining computational models with intracellular recordings of neocortical pyramidal cells in vivo, we provided a plausible explanation for our experimental observations on how action potential are controlled by IPSPs in these cells. odels and experiments suggest that IPSPs aect action potentials bytwo mechanisms: ashunt eect due to the opening of ion channels underlying the IPSPs, and a voltagedependent eect by preventing dendritic Na + channels to participate to the somatic spike. Further, we suggest that under conditions of synaptic activity that occurs during active states in vivo, the conductance shunt and the voltage-dependent eect of synaptic inputs do not provide favorable conditions for backpropagating action potentials 17. REFERENCES 1. Stuart, G., Spruston, N., Sakmann, B. and Hausser,. Action potential initiation and backpropagation in neurons of the mammalian CNS. Trends Neurosci. 20: 125-131 (1997). 2. Kim H.G., Beierlein. and Connors B.W. Inhibitory control of excitable dendrites in neocortex. J. Neurophysiol. 74: 1810-1814 (1995). 3. Tsubokawa H. and Ross W.N. IPSPs modulate spike backpropagation and associated [Ca 2+ ] changes in the dendrites of hippocampal CA1 pyramidal neurons. J. Neurophysiol. 76: 2896-2906 (1996). 4. Pare,., Lang, E.J. and estexhe, A. Inhibitory control of somatic and dendritic sodium spikes in neocortical pyramidal neurons in vivo: anintracellular and computational study. Neuroscience, in press (1998). 5. Lang, E.J. and Pare,. Synaptic and synaptically activated intrinsic conductances underlie inhibitory potentials in cat lateral amygdaloid projection neurons in vivo. J. Neurophysiol. 77: 353-363 (1997) 6. agee J.C. and Johnston. Characterization of single voltage-gated Na + and Ca 2+ channels in apical dendrites of rat CA1 pyramidal neurons. J. Physiol. 487: 67-90 (1995). 7. Black J.A., Kocsis J.. and Waxman S.G. Ion channel organization of the myelinated ber. Trends Neurosci. 13: 48-54 (1990). 8. ainen Z.F., Joerges J., Huguenard J.R. and Sejnowski T.J. A model of spike initiation in neocortical pyramidal neurons. Neuron 15: 1427-1439 (1995). 9. Rapp,., Yarom, Y., and Segev, I. odeling back propagating action potential in weakly excitable dendrites of neocortical pyramidal cells. Proc. Natl. Acad. Sci. USA 93: 11985-11990 (1996). 10. Traub R.. and iles R. Neuronal Networks of the Hippocampus. Cambridge University Press, Cambridge (1991). 11. efelipe J. and Fari~nas I. The pyramidal neuron of the cerebral cortex: morphological and chemical characteristics of the synaptic inputs. Progress Neurobiol. 39: 563-607 (1992). 12. White E. L. Cortical circuits. Birkhauser, Boston, A (1989). 13. estexhe, A., ainen, Z.F. and Sejnowski, T.J. Kinetic models of synaptic transmission. In: ethods in Neuronal odeling (2nd ed), edited by C. Koch and I. Segev. IT Press, Cambridge, A, pp. 1-26 (1998). 14. Callaway J.C. and Ross W.N. Frequency-dependent propagation of sodium action potentials in dendrites of hippocampal CA1 pyramidal neurons. J. Neurophysiol. 74: 1395-1403 (1995). 15. Spruston N., Schiller Y., Stuart G. and Sakmann B. Activity-dependent action potential invasion and calcium inux into hippocampal CA1 dendrites. Science 268: 297-300 (1995). 16. Svoboda, K., enk, W., Kleinfeld,. and Tank,.W. In vivo calcium dynamics in neocortical pyramidal neurons. Nature 385: 161-165 (1997). 17. Research supported by grants from RC, NSERC and FRSQ. E.J.L. was supported by a NINS fellowship.